insns.pl

汇编编译器的最新版本的源码.买了自己动手写操作系统这本书的人一定要下

#!/usr/bin/perl
#
# insns.pl   produce insnsa.c, insnsd.c, insnsi.h, insnsn.c from insns.dat
#
# The Netwide Assembler is copyright (C) 1996 Simon Tatham and
# Julian Hall. All rights reserved. The software is
# redistributable under the license given in the file "LICENSE"
# distributed in the NASM archive.

# Opcode prefixes which need their own opcode tables
# LONGER PREFIXES FIRST!
@disasm_prefixes = qw(0F24 0F25 0F38 0F3A 0F7A 0FA6 0FA7 0F);

# This should match MAX_OPERANDS from nasm.h
$MAX_OPERANDS = 5;

# Add VEX prefixes
@vexlist = ();
for ($m = 0; $m < 32; $m++) {
    for ($lp = 0; $lp < 8; $lp++) {
	push(@vexlist, sprintf("VEX%02X%01X", $m, $lp));
    }
}
@disasm_prefixes = (@vexlist, @disasm_prefixes);

@bytecode_count = (0) x 256;

print STDERR "Reading insns.dat...\n";

@args   = ();
undef $output;
foreach $arg ( @ARGV ) {
    if ( $arg =~ /^\-/ ) {
	if  ( $arg =~ /^\-([abdin])$/ ) {
	    $output = $1;
	} else {
	    die "$0: Unknown option: ${arg}\n";
	}
    } else {
	push (@args, $arg);
    }
}

$fname = "insns.dat" unless $fname = $args[0];
open (F, $fname) || die "unable to open $fname";

%dinstables = ();
@bytecode_list = ();

$line = 0;
$insns = 0;
while (<F>) {
  $line++;
  chomp;
  next if ( /^\s*(\;.*|)$/ );   # comments or blank lines

  unless (/^\s*(\S+)\s+(\S+)\s+(\S+|\[.*\])\s+(\S+)\s*$/) {
      warn "line $line does not contain four fields\n";
      next;
  }
  @fields = ($1, $2, $3, $4);
  ($formatted, $nd) = format_insn(@fields);
  if ($formatted) {
    $insns++;
    $aname = "aa_$fields[0]";
    push @$aname, $formatted;
  }
  if ( $fields[0] =~ /cc$/ ) {
      # Conditional instruction
      $k_opcodes_cc{$fields[0]}++;
  } else {
      # Unconditional instruction
      $k_opcodes{$fields[0]}++;
  }
  if ($formatted && !$nd) {
    push @big, $formatted;
    my @sseq = startseq($fields[2]);
    foreach $i (@sseq) {
	if (!defined($dinstables{$i})) {
	    $dinstables{$i} = [];
	}
	push(@{$dinstables{$i}}, $#big);
    }
  }
}

close F;

#
# Generate the bytecode array.  At this point, @bytecode_list contains
# the full set of bytecodes.
#

# Sort by descending length
@bytecode_list = sort { scalar(@$b) <=> scalar(@$a) } @bytecode_list;
@bytecode_array = ();
%bytecode_pos = ();
$bytecode_next = 0;
foreach $bl (@bytecode_list) {
    my $h = hexstr(@$bl);
    next if (defined($bytecode_pos{$h}));

    push(@bytecode_array, $bl);
    while ($h ne '') {
	$bytecode_pos{$h} = $bytecode_next;
	$h = substr($h, 2);
	$bytecode_next++;
    }
}
undef @bytecode_list;

@opcodes    = sort keys(%k_opcodes);
@opcodes_cc = sort keys(%k_opcodes_cc);

if ( !defined($output) || $output eq 'b') {
    print STDERR "Writing insnsb.c...\n";

    open B, ">insnsb.c";

    print B "/* This file auto-generated from insns.dat by insns.pl" .
        " - don't edit it */\n\n";

    print B "#include \"nasm.h\"\n";
    print B "#include \"insns.h\"\n\n";

    print B "const uint8_t nasm_bytecodes[$bytecode_next] = {\n";

    $p = 0;
    foreach $bl (@bytecode_array) {
	printf B "    /* %5d */ ", $p;
	foreach $d (@$bl) {
	    printf B "%#o,", $d;
	    $p++;
	}
	printf B "\n";
    }
    print B "};\n";

    print B "\n";
    print B "/*\n";
    print B " * Bytecode frequencies (including reuse):\n";
    print B " *\n";
    for ($i = 0; $i < 32; $i++) {
	print B " *";
	for ($j = 0; $j < 256; $j += 32) {
	    print B " |" if ($j);
	    printf B " %3o:%4d", $i+$j, $bytecode_count[$i+$j];
	}
	print B "\n";
    }
    print B " */\n";

    close B;
}

if ( !defined($output) || $output eq 'a' ) {
    print STDERR "Writing insnsa.c...\n";

    open A, ">insnsa.c";

    print A "/* This file auto-generated from insns.dat by insns.pl" .
        " - don't edit it */\n\n";

    print A "#include \"nasm.h\"\n";
    print A "#include \"insns.h\"\n\n";

    foreach $i (@opcodes, @opcodes_cc) {
	print A "static const struct itemplate instrux_${i}[] = {\n";
	$aname = "aa_$i";
	foreach $j (@$aname) {
	    print A "    ", codesubst($j), "\n";
	}
	print A "    ITEMPLATE_END\n};\n\n";
    }
    print A "const struct itemplate * const nasm_instructions[] = {\n";
    foreach $i (@opcodes, @opcodes_cc) {
	print A "    instrux_${i},\n";
    }
    print A "};\n";

    close A;
}

if ( !defined($output) || $output eq 'd' ) {
    print STDERR "Writing insnsd.c...\n";

    open D, ">insnsd.c";

    print D "/* This file auto-generated from insns.dat by insns.pl" .
        " - don't edit it */\n\n";

    print D "#include \"nasm.h\"\n";
    print D "#include \"insns.h\"\n\n";

    print D "static const struct itemplate instrux[] = {\n";
    $n = 0;
    foreach $j (@big) {
	printf D "    /* %4d */ %s\n", $n++, codesubst($j);
    }
    print D "};\n";

    foreach $h (sort(keys(%dinstables))) {
	next if ($h eq '');	# Skip pseudo-instructions
	print D "\nstatic const struct itemplate * const itable_${h}[] = {\n";
	foreach $j (@{$dinstables{$h}}) {
	    print D "    instrux + $j,\n";
	}
	print D "};\n";
    }

    @prefix_list = ();
    foreach $h (@disasm_prefixes, '') {
	for ($c = 0; $c < 256; $c++) {
	    $nn = sprintf("%s%02X", $h, $c);
	    if ($is_prefix{$nn} || defined($dinstables{$nn})) {
		# At least one entry in this prefix table
		push(@prefix_list, $h);
		$is_prefix{$h} = 1;
		last;
	    }
	}
    }

    foreach $h (@prefix_list) {
	print D "\n";
	print D "static " unless ($h eq '');
	print D "const struct disasm_index ";
	print D ($h eq '') ? 'itable' : "itable_$h";
	print D "[256] = {\n";
	for ($c = 0; $c < 256; $c++) {
	    $nn = sprintf("%s%02X", $h, $c);
	    if ($is_prefix{$nn}) {
		die "$fname: ambiguous decoding of $nn\n"
		    if (defined($dinstables{$nn}));
		printf D "    { itable_%s, -1 },\n", $nn;
	    } elsif (defined($dinstables{$nn})) {
		printf D "    { itable_%s, %u },\n",
		$nn, scalar(@{$dinstables{$nn}});
	    } else {
		printf D "    { NULL, 0 },\n";
	    }
	}
	print D "};\n";
    }

    print D "\nconst struct disasm_index * const itable_VEX[32][8] = {\n   ";
    for ($m = 0; $m < 32; $m++) {
	print D " {\n";
	for ($lp = 0; $lp < 8; $lp++) {
	    $vp = sprintf("VEX%02X%01X", $m, $lp);
	    if ($is_prefix{$vp}) {
		printf D "        itable_%s,\n", $vp;
	    } else {
		print  D "        NULL,\n";
	    }
	}
	print D "    },";
    }
    print D "\n};\n";

    close D;
}

if ( !defined($output) || $output eq 'i' ) {
    print STDERR "Writing insnsi.h...\n";

    open I, ">insnsi.h";

    print I "/* This file is auto-generated from insns.dat by insns.pl" .
        " - don't edit it */\n\n";
    print I "/* This file in included by nasm.h */\n\n";

    print I "/* Instruction names */\n\n";
    print I "#ifndef NASM_INSNSI_H\n";
    print I "#define NASM_INSNSI_H 1\n\n";
    print I "enum opcode {\n";
    $maxlen = 0;
    foreach $i (@opcodes, @opcodes_cc) {
	print I "\tI_${i},\n";
	$len = length($i);
	$len++ if ( $i =~ /cc$/ ); # Condition codes can be 3 characters long
	$maxlen = $len if ( $len > $maxlen );
    }
    print I "\tI_none = -1\n";
    print I "\n};\n\n";
    print I "#define MAX_INSLEN ", $maxlen, "\n";
    print I "#define FIRST_COND_OPCODE I_", $opcodes_cc[0], "\n\n";
    print I "#endif /* NASM_INSNSI_H */\n";

    close I;
}

if ( !defined($output) || $output eq 'n' ) {
    print STDERR "Writing insnsn.c...\n";

    open N, ">insnsn.c";

    print N "/* This file is auto-generated from insns.dat by insns.pl" .
        " - don't edit it */\n\n";
    print N "#include \"tables.h\"\n\n";

    print N "const char * const nasm_insn_names[] = {";
    $first = 1;
    foreach $i (@opcodes, @opcodes_cc) {
	print N "," if ( !$first );
	$first = 0;
	$ilower = $i;
	$ilower =~ s/cc$//;	# Remove conditional cc suffix
	$ilower =~ tr/A-Z/a-z/;	# Change to lower case (Perl 4 compatible)
	print N "\n\t\"${ilower}\"";
    }
    print N "\n};\n";
    close N;
}

printf STDERR "Done: %d instructions\n", $insns;

# Count primary bytecodes, for statistics
sub count_bytecodes(@) {
    my $skip = 0;
    foreach my $bc (@_) {
	if ($skip) {
	    $skip--;
	    next;
	}
	$bytecode_count[$bc]++;
	if ($bc >= 01 && $bc <= 03) {
	    $skip = $bc;
	} elsif (($bc & ~03) == 010) {
	    $skip = 1;
	} elsif (($bc & ~013) == 0144) {
	    $skip = 1;
	} elsif ($bc == 0172) {
	    $skip = 1;
	} elsif ($bc >= 0260 && $bc <= 0270) {
	    $skip = 2;
	} elsif ($bc == 0330) {
	    $skip = 1;
	}
    }
}

sub format_insn(@) {
    my ($opcode, $operands, $codes, $flags) = @_;
    my $num, $nd = 0;
    my @bytecode;

    return (undef, undef) if $operands eq "ignore";

    # format the operands
    $operands =~ s/:/|colon,/g;
    $operands =~ s/mem(\d+)/mem|bits$1/g;
    $operands =~ s/mem/memory/g;
    $operands =~ s/memory_offs/mem_offs/g;
    $operands =~ s/imm(\d+)/imm|bits$1/g;
    $operands =~ s/imm/immediate/g;
    $operands =~ s/rm(\d+)/rm_gpr|bits$1/g;
    $operands =~ s/(mmx|xmm|ymm)rm/rm_$1/g;
    $operands =~ s/\=([0-9]+)/same_as|$1/g;
    if ($operands eq 'void') {
	@ops = ();
    } else {
	@ops = split(/\,/, $operands);
    }
    $num = scalar(@ops);
    while (scalar(@ops) < $MAX_OPERANDS) {
	push(@ops, '0');
    }
    $operands = join(',', @ops);
    $operands =~ tr/a-z/A-Z/;

    # format the flags
    $flags =~ s/,/|IF_/g;
    $flags =~ s/(\|IF_ND|IF_ND\|)//, $nd = 1 if $flags =~ /IF_ND/;
    $flags = "IF_" . $flags;

    @bytecode = (decodify($codes), 0);
    push(@bytecode_list, [@bytecode]);
    $codes = hexstr(@bytecode);
    count_bytecodes(@bytecode);

    ("{I_$opcode, $num, {$operands}, \@\@CODES-$codes\@\@, $flags},", $nd);
}

#
# Look for @@CODES-xxx@@ sequences and replace them with the appropriate
# offset into nasm_bytecodes
#
sub codesubst($) {
    my($s) = @_;
    my $n;

    while ($s =~ /\@\@CODES-([0-9A-F]+)\@\@/) {
	my $pos = $bytecode_pos{$1};
	if (!defined($pos)) {
	    die "$fname: no position assigned to byte code $1\n";
	}
	$s = $` . "nasm_bytecodes+${pos}" . "$'";
    }